Every Claude Code cheat sheet I bookmarked was outdated within a month.
So I made a verified one.
Getting the Most Out of Claude Code
Verified July 2026 (CLI v2.1.211) · Compiled by Douglas Mun
Claude Code is an agentic coding assistant — it reads your repo, runs commands, and edits files within the permission boundaries you set. The gap between casual and expert use is narrow but compounds: experts plan before they build, give Claude durable memory, isolate risky work in worktrees, and offload repetition to skills and subagents.
How to read this. Section colours mark stable features. new entries arrived in 2026 (v2.1.84+). Struck-through entries are removed. Use the HTML version's search box to filter instantly.
Path to expert. Start in plan mode (Shift+Tab or /plan) for anything multi-file. Put project context in CLAUDE.md at your repo root — it survives compaction. Codify repeated workflows as skills in .claude/skills/. Isolate risky work with --worktree. Use /btw for side questions. Raise to /effort high or xhigh for hard problems on Opus 4.8. Install an LSP plugin so Claude sees type errors as it edits.
Version 2.0 — what should I add next?